From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <axboe@kernel.dk>
Message-ID: <530CCDE3.80109@kernel.dk>
Date: Tue, 25 Feb 2014 09:07:47 -0800
From: Jens Axboe <axboe@kernel.dk>
MIME-Version: 1.0
Subject: Re: numerical option parsing broken in engine options
References: <CALKLVuJOou3sLCgnKyfpUaB52M_4L+Xq6_U7ft1YFss_B-+xoQ@mail.gmail.com>
In-Reply-To: <CALKLVuJOou3sLCgnKyfpUaB52M_4L+Xq6_U7ft1YFss_B-+xoQ@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
To: Castor Fu <castor@egocast.org>
Cc: fio@vger.kernel.org
List-ID: <fio@vger.kernel.org>

On 2014-02-24 17:51, Castor Fu wrote:
> I was debugging some problems with parsing fio jobs, and found
> options.c:fio_get_kb_base is broken for external engines... The
> parameter
> 'data' is either 0, td, or td->eo.  If it's td->eo it's not really
> going to work.
>
> The cleaner fix is probably to unwind passing 'data' so far down the
> stack, but it
> touches a lot of functions, so I thought I'd send this out now.
>
> unsigned int fio_get_kb_base(void *data)
> {
>          struct thread_options *o = data;
>          unsigned int kb_base = 0;
>
>          if (o)
>                  kb_base = o->kb_base;
>          if (!kb_base)
>                  kb_base = 1024;
>
>          return kb_base;
> }
>
>
> If I replace
>     if(o)
> by
>       if (o && (data - (void *) threads) % sizeof(threads[0]) == 0)
>
> then it behaves as I expect when engine options use the 'k', etc.

Not sure I follow... Are you calling fio_get_kb_base() from the external 
engine?

-- 
Jens Axboe